AI consultant

What an hour actually produces

You describe the work that repeats in your business. I ask about volume, who does it now, and what a mistake costs. By the end you have three things, in plain words, whether or not we ever work together.

• Which one job is worth automating first, and why that one before the others.
• A rough read on what it is worth in hours or money, using your own numbers, not a benchmark.
• What has to be true before it can be built at all — usually a data problem, not a model problem.

Where AI is the wrong answer

This is the part a consultancy paid by the day rarely says out loud, so it goes first.

• The task happens twice a month. Automation earns its keep on repetition. If a person does it twice a month in twenty minutes, the build will cost more than the job for years.
• Nobody can say what the right answer looks like. If two people in your business would judge the same case differently, a machine has nothing to learn from. Settle the rule between people first; then it is worth automating.
• The data does not exist yet. You cannot forecast, score or route on numbers nobody has been recording. Sometimes the honest first project is a boring pipeline, and it is worth doing on its own before any AI runs on top.
• The mistake is expensive and invisible. If a wrong output costs real money and nobody would notice for a month, the automation needs a person in the loop — which changes the economics, and sometimes kills the case.

What the hour actually looks like

No slides and no discovery phase. You talk, I ask, and the questions are always the same four because they are the ones that decide everything: how often does this happen, who does it now, what does one mistake cost, and where do the numbers live.

Most of the hour goes on the last one. It is where the answer usually changes — a job that sounded like a model problem turns out to be two systems that disagree about the same record, and that is a different, cheaper, more useful project.

You leave with the first job named, a rough sense of what it is worth in your own numbers, and the thing that has to be true before it can be built. In writing, the same day.

How the work runs after that

• You talk to me, not to an account manager. There is no layer of people between you and the person building it. What carries the volume is a team of AI systems doing the heavy lifting, which is the whole reason one person can.
• The first piece works in about two weeks. One job, end to end, on real data, where you can watch it run on a Monday — not a pilot in a sandbox that quietly stops in March.
• You own it. What gets built stays yours, logins included. No lock-in, and no per-seat fee on something you already paid for.

Who this is not for

• You want an AI strategy document for a board. I do not write those.
• You want training or a course. I do not teach; I build.
• You want a headcount reduction plan. AI here takes the routine and the round-the-clock grind; the people stay and get more done. If that is not what you are after, we are not a fit.